About how to ensure anonymity of data across various releases
Have k-anonymity as the basis, based on end use-case either prefer global or local recoding
No records are deleted, in the subsequent releases, the cumulative summation or the union of all the past records are along with the present data inputs are considered for data release
Highlights, Backward, forward, cross attack and ways to overcome
Temporal attacks are referred to as correspondence attacks, i.e how are you able to relate 2 or more tables
Instantiation mapping or making buddies is basically the process of mapping records in table R1 to those in table R2, where R1 is the earlier table published at timestamp T1 and R2 is the table published at time T2, given as R2 = R2 U R1, i.e the history is not deleted . whatever is the case it is added and made as union
In R2, all records of R1 will have a mapping as R2 = R2 U R1, and in R2 there will be exactly R1 number of records that have a mapping
Way to identify B, C, F attack and prevent its occurrence, this is what it covers next
Talks about crack size → read once
Our attack model is based on excluding such non-representing records